Question

我有一个非常大的数据库（大约有3000万条记录，每条记录至少包含26个字段），我已经使用Apache Lucene Java编制索引。

我正在构建一个来自两个字段的查询。每个搜索词都可以出现在九个字段中的任何一个中，如果两个搜索词出现在Document中的任何相关字段中，我希望我的查询返回一个Document。查询的结构如下：

Private Query CreateQuery(String theSearchTerm, String theField) throws ParseException
{
    StandardAnalyzer theAnalyzer = new StandardAnalyzer(Version.LUCENE_35);
    Query q;
    QueryParser qp = new QueryParser(Version.LUCENE_35, theField, theAnalyzer);
    qp.setDefaultOperator(QueryParser.Operator.AND);
    qp.setAllowLeadingWildcard = true;
    q = qp.parse(theSearchTerm);
    return q;
}

Public ScoreDoc[] RunTheQuery(String searchTerm1, String searchTerm2)
{
    Directory theIndex = new SimpleFSDirectory(new File("C:\\MyDirectory");
    IndexSearcher theSearcher = new IndexSearcher(InderReader.open(theIndex));

    BooleanQuery theTopLevelBooleanQuery = new BooleanQuery();

    BooleanQuery fields1 = new BooleanQuery();
    BooleanQuery fields2 = new BooleanQuery();
    BooleanQuery fields3 = new BooleanQuery();
    BooleanQuery fields4 = new BooleanQuery();
    BooleanQuery fields5 = new BooleanQuery();
    BooleanQuery fields6 = new BooleanQuery();
    BooleanQuery fields7 = new BooleanQuery();
    BooleanQuery fields8 = new BooleanQuery();
    BooleanQuery fields9 = new BooleanQuery();

    BooleanQuery innerQuery = new BooleanQuery();

    fields1.add(CreateQuery(searchTerm1, param1), BooleanClause.Occur.MUST);
    fields1.add(CreateQuery(searchTerm2, param2), BooleanClause.Occur.MUST);
    fields2.add(CreateQuery(searchTerm1, param3), BooleanClause.Occur.MUST);
    fields2.add(CreateQuery(searchTerm2, param4), BooleanClause.Occur.MUST);
    fields3.add(CreateQuery(searchTerm1, param5), BooleanClause.Occur.MUST);
    fields3.add(CreateQuery(searchTerm2, param6), BooleanClause.Occur.MUST);
    fields4.add(CreateQuery(searchTerm1, param7), BooleanClause.Occur.MUST);
    fields4.add(CreateQuery(searchTerm2, param8), BooleanClause.Occur.MUST);
    fields5.add(CreateQuery(searchTerm1, param9), BooleanClause.Occur.MUST);
    fields5.add(CreateQuery(searchTerm2, param10), BooleanClause.Occur.MUST);
    fields6.add(CreateQuery(searchTerm1, param11), BooleanClause.Occur.MUST);
    fields6.add(CreateQuery(searchTerm2, param12), BooleanClause.Occur.MUST);
    fields7.add(CreateQuery(searchTerm1, param13), BooleanClause.Occur.MUST);
    fields7.add(CreateQuery(searchTerm2, param14), BooleanClause.Occur.MUST);
    fields8.add(CreateQuery(searchTerm1, param15), BooleanClause.Occur.MUST);
    fields8.add(CreateQuery(searchTerm2, param16), BooleanClause.Occur.MUST);
    fields9.add(CreateQuery(searchTerm1, param17), BooleanClause.Occur.MUST);
    fields9.add(CreateQuery(searchTerm2, param18), BooleanClause.Occur.MUST);

    innerQuery.add(fields1, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields2, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields3, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields4, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields5, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields6, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields7, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields8, BooleanClause.Occur.SHOULD);
    innerQuery.add(fields9, BooleanClause.Occur.SHOULD);

    theTopLevelBooleanQuery.add(innerQuery, BooleanClause.Occur.MUST);

    TopDocScoreCollector collector = TopDocScoreCollector.create(200, true);

    //Heap space error occurs here
    theSearcher.search(theTopLevelBooleanQuery, collector);

    ScoreDoc[] hits = collector.topDocs().scoreDocs;
    return hits;
}

我的问题是，当我调用IndexSearcher.search（）方法时，服务器上的java.exe进程（Windows Server 2003 R2）消耗超过540 MB，这会导致java堆空间错误。为了完整起见，java应用程序在Web服务器上运行（目前是Oracle Glassfish，虽然我希望转移到Apache Tomcat）。

有没有人知道如何停止此堆空间错误？ StackOverflow帖子（http://stackoverflow.com/questions/7259736/cant-open-lucene-index-java-heap-space）似乎解决了类似的问题，但并没有给出详细的答案。

是增加Java进程可以使用的内存量的唯一答案吗？是写一个新搜索者的唯一答案，在这种情况下，任何人都可以推荐一篇关于轻量级搜索者的好文章吗？

有没有办法通过修改上面的代码来解决这个问题？

感谢任何帮助，谢谢，瑞克

Answer 1

您可以像这样增加java堆空间：

java -Xmx6g myprogram

或看到这篇文章： increase heap size in Java

或： IBM SDK for Java

Lucene：IndexSearcher.search（）会在非常大的数据库上导致java堆空间错误

1 个答案: